What is an online feature store?
An online feature store is the part that serves fresh feature values to a model at request time. It holds the current value of each feature, ready to read within the latency budget of a live prediction.
Online vs offline serving
The offline half of a feature store computes and stores history for training. The online half serves the current value of each feature to a model when it runs. Online serving is defined by a budget: the value has to come back inside the time a live request can wait.
What it takes to serve online
Serving online means keeping each feature’s current value computed and ready, not calculated on demand from raw data. Something has to update those values as new events arrive, and store them where they can be read quickly. The fresher the update, the more recent the behavior the model can act on.
Why it is the harder half
Computing features in batch is well understood. Serving them live, kept current as behavior happens, within a request budget, is the part teams underestimate. It is a streaming and serving problem on top of the modeling one, and it is where an online feature store earns its place.
See it on your own traffic
Add the SDK, define one attribute, and see an attribute updating against your own traffic in the first session. The free tier runs on the real engine: no card, no sales call.